Solar cell temperature prediction model of support vector machine optimized by particle swarm optimization algorithm

Abstract

Establishing a general and precise solar cell temperature model is of crucial importance for photovoltaic system modeling, the loss analysis of output power, and conversion efficiency. According to the complex mechanism of solar cell temperature, in this paper we study the steady state thermal model (SSTM) of solar cell temperature and accurate prediction model of method of support vector machine (SVM). Firstly, based on the approximate linear relationship among air temperature, solar radiation intensity, wind speed and solar cell temperature, the polynomial model of solar cell temperature is established and the unknown parameters of the model are extracted with the improved differential evolution algorithm. Secondly, in order to improve the accuracy of SVM prediction model, the particle swarm optimization algorithm is adopted to optimize the parameters (including kernel parameter g and penalty factor C from the radial basis function kernel) of SVM. After the input/output sample set is determined and the training set and test set are classified, a prediction model of solar cell temperature based on particle swarm optimization support vector machine is established. Finally, experimental acquisition platform is built to reduce the influences of air humidity, solar incidence angle, and thermal hysteresis effects on PV cell temperature. Through contrasting experiments, it is shown that the established fitting of the SSTM is better than the models given in other literature, and the prediction model is reliable, comprehensive and simple. The selected parameter optimization algorithm is superior to genetic algorithm and cross-validation method established on the optimization performance, and the accuracy of prediction model is superior to the prediction performance of back propagation neural network and identified SSTM.